(I definitely should have thought of this earlier; interestingly enough it was this comment that was the trigger.)
Use probabilities! (Or likelihood ratios.) Especially when arguing. Yes, do so with care, i.e. without deceiving yourself into thinking you’re better calibrated than you are—but hiding the fact that you’re not perfectly calibrated doesn’t make your calibration any better. You brain is still making the same mistakes whether you choose to make them verbally explicit or not. So instead of reacting with indignation when someone disagrees, just ask “how confident are you?” Then, if necessary, follow up with “what was your prior?” “How many bits of evidence is this piece of information worth?” Make your argument a numerical game rather than a barefaced status war.
People often profess to be uncomfortable with assigning a numerical value to their confidence level. I have found, strangely enough, that switching from percentages (or values between 0 and 1) to something that feels more “discrete”, like a scale of 0 to 10 or a “five-star” system, sometimes helps with this.
I have found, strangely enough, that switching from percentages (or values between 0 and 1) to something that feels more “discrete”, like a scale of 0 to 10 or a “five-star” system, sometimes helps with this.
This makes perfect sense to me. I feel far more comfortable converting my sense of credibility to an intensity scale of 1 to 100 than converting those intensities to probabilities.
I have found, strangely enough, that switching from percentages (or values between 0 and 1) to something that feels more “discrete”, like a scale of 0 to 10 or a “five-star” system, sometimes helps with this.
Have you considered using decibans for this purpose?
(I definitely should have thought of this earlier; interestingly enough it was this comment that was the trigger.)
Use probabilities! (Or likelihood ratios.) Especially when arguing. Yes, do so with care, i.e. without deceiving yourself into thinking you’re better calibrated than you are—but hiding the fact that you’re not perfectly calibrated doesn’t make your calibration any better. You brain is still making the same mistakes whether you choose to make them verbally explicit or not. So instead of reacting with indignation when someone disagrees, just ask “how confident are you?” Then, if necessary, follow up with “what was your prior?” “How many bits of evidence is this piece of information worth?” Make your argument a numerical game rather than a barefaced status war.
People often profess to be uncomfortable with assigning a numerical value to their confidence level. I have found, strangely enough, that switching from percentages (or values between 0 and 1) to something that feels more “discrete”, like a scale of 0 to 10 or a “five-star” system, sometimes helps with this.
This makes perfect sense to me. I feel far more comfortable converting my sense of credibility to an intensity scale of 1 to 100 than converting those intensities to probabilities.
Have you considered using decibans for this purpose?